Source Apportionment of Annual Water Pollution Loads in River Basins by Remote-Sensed Land Cover Classiﬁcation

water

Article Source Apportionment of Annual Water Pollution Loads in River Basins by Remote-Sensed Land Cover Classiﬁcation

Yi Wang 1, Bin He 2,*, Weili Duan 2,*, Weihong Li 1, Pingping Luo 3,4 and Bam H. N. Razaﬁndrabe 5

1 State Key Laboratory of Desert and Oasis Ecology, Xinjiang Institute of Ecology and Geography, Chinese Academy of Sciences, Urumqi 830011, China; [email protected] (Y.W.); [email protected] (W.L.) 2 Key Laboratory of Watershed Geographic Science, Nanjing Institute of Geography and Limnology, Chinese Academy of Sciences, Nanjing 210008, China 3 Key Laboratory of Subsurface Hydrology and Ecological Effects in Arid Region (Chang’an University), Ministry of Education, Xi’an 710064, China; [email protected] 4 School of Environmental Science and Engineering, Chang’an University, Xi’an 710064, China 5 Faculty of Agriculture, University of the Ryukyus, Nishihara, Okinawa 903-0213, Japan; [email protected] * Correspondence: [email protected] (B.H.), [email protected] (W.D.); Tel.: +86-025-8688-2171 (B.H.); +86-025-8688-2173 (W.D.)

Academic Editor: Y. Jun Xu Received: 4 April 2016; Accepted: 9 August 2016; Published: 23 August 2016

Abstract: In this study, in order to determine the efﬁciency of estimating annual water pollution loads from remote-sensed land cover classiﬁcation and ground-observed hydrological data, an empirical model was investigated. Remote sensing data imagery from National Oceanic and Atmospheric Administration (NOAA) Advanced Very High Resolution Radiometer were applied to an 11 year (1994–2004) water quality dataset for 30 different rivers in Japan. Six water quality indicators—total nitrogen (TN), total phosphorus (TP), biochemical oxygen demand (BOD), chemical oxygen demand (COD), and dissolved oxygen (DO)—were examined by using the observed river water quality data and generated land cover map. The TN, TP, BOD, COD, and DO loads were estimated for the 30 river basins using the empirical model. Calibration (1994–1999) and validation (2000–2004) results showed that the proposed simulation technique was useful for predicting water pollution loads in the river basins. We found that vegetation land cover had a larger impact on TP export into all rivers. Urban areas had a very small impact on DO export into rivers, but a relatively large impact on BOD and TN export. The results indicate that the application of land cover data generated from the remote-sensed imagery could give a useful interpretation about the river water quality.

Keywords: Japan; river; water quality; remote sensing; AVHRR

1. Introduction Changing land use and land management practices are regarded as amongst the most important factors that can alter hydrological systems and water quality, which have become increasingly important to catchment stakeholders, such as management groups, land owners and government departments [1–4]. Generally, a river’s water quality is linked to the land cover in the watershed and is degraded by changes in the land cover patterns on their watersheds as human activities increase [5–8]. Different studies have increasingly recognized that human action at the landscape scale is a principal threat to the ecological integrity of river ecosystems and water quality [9–15]. However, the information on the sources of pollutants in catchments and on the response of water quality to changing land

Water 2016, 8, 361; doi:10.3390/w8090361 www.mdpi.com/journal/water Water 2016, 8, 361 2 of 14 use practices is still limited in many catchments [16–20]. It is therefore important to understand the relationships between catchment characteristics and river water chemistry, which provides a base for determining how future changes in land cover and use and climate will impact on river water quality. In Japan, the water quality of rivers has been improved by legislation on sewage, drainage, etc. However, the water quality cannot be said to have been restored to its natural conditions because the pollution from non-point sources can increase even when factory drainage and domestic drainage are improved by law [21]. Therefore, it is still necessary to take prevention measures to control non-point source pollutants for maintaining the natural ecosystem and environment [22,23]. Among these, identiﬁcation of the water pollution loads from different sources is important. To this end, statistic methods and hydrological modeling have been proposed and applied to estimate the contribution from different water pollution sources. By using multivariate statistical techniques, Shrestha and Kazama [24] evaluated the temporal/spatial variations in the Fuji river basin, illustrating the usefulness of multivariate statistical techniques for identifying pollution sources/factors and understanding temporal/spatial variation. Duan et al. (2015) developed a SPARROW-based (SPAtially Referenced Regression on Watershed Attributes) watershed model to estimate the sources and transport of suspended sediments in surface waters of the Ishikari River basin [25]. Recently, remote sensing was used to evaluating water quality. Oki and Yasuoka [26] mapped the potential annual total nitrogen load in the river basins of Japan with remotely sensed imagery of 2006. However, the non-point source pollutants are transported to water body is still unclear. Therefore, in a previous study we have analyzed the relationship between land cover types and potential annual water pollution loads and improved an empirical model to successfully calculate potential annual water pollution loads in 30 river basins in Japan by using the collected dataset in the year of 1996 [27]. Based on these previous results of estimating total water pollution loads in year 1996 for 30 river basins, the objectives of the present study are to (1) test the model prediction capability with long term dataset; (2) modify the original empirical model to account for the water pollution loads from each land cover classiﬁcation; (3) examine and analyze the capability of this method for apportioning long term potential annual loads of water quality indicators such as total nitrogen (TN), total phosphorus (TP), biochemical oxygen demand (BOD), chemical oxygen demand (COD), and dissolved oxygen (DO) in river basins. The results of this study will be very useful to model the linkage between river water quality and land cover classes by separately considering the impact of each land cover type on the water pollution loads.

2. Materials and Methods

2.1. Collection of Long-Term Dataset In order to measure and assess changes in vegetation phenology and conditions as well as perform land cover type classiﬁcation, this study uses 26 National Oceanic and Atmospheric Administration (NOAA) Advanced Very High Resolution Radiometer (AVHRR) images showing maximum annual normalized difference vegetation index (NDVI), at 1.1 km pixel resolution, across Japan for each year between 1984 and 2009. The NOAA AVHRR data was produced by the Sawada and Takeuchi Laboratory, Institute of Industrial Science, University of Tokyo. They were used to produce a land cover distribution map [28]. AVHRR data were geometrically corrected based on ground control point (GCP) matching by using PaNDA software, and registration error over the image was less than 1 pixel. PaNDA is a free software package for NOAA data analysis [29]. The AVHRR NDVI products were also provided by laboratory of Sawada and Takeuchi [28]. The NDVI can be calculated according to the following equation [30]: (ρ − ρ ) NDVI = 780 670 (1) (ρ780 + ρ670) where, ρ780 represents the near infrared band value for a cell, ρ670 represents the red band value for the cell. Daily NDVI was calculated using Band 1 (0.58–0.68 µm) and Band 2 (0.725–1.10 µm) Water 2016, 8, 361 3 of 14 images to produce the monthly maximum NDVI imagery, and by using the 12 monthly maximum NDVI imagery datasets, the effects of clouds on seasonal land cover changes can be removed to produce each yearly maximum images. All the processing was conducted by using the remote sensing softwareWater 2016, ERDAS8, 361 IMAGE (Version 9.2). The Iterative Self-Organizing Data Analysis3 of 13 Technique

Algorithmimages (ISODATA) to produce [31 the,32 monthly], which maximum is an unsupervised NDVI imagery, classification, and by using the was 12 monthly used to maximum produce a yearly land coverNDVI classification imagery datasets, map the of effects Japan of using clouds theon seasonal 12 monthly land cover maximum changes can NDVI be removed imagery to datasets. Figure1 showsproduce an each example yearly maximum of the generated images. All landthe processing cover maps was conducted in 1996. by The using river the basin remote map was superimposedsensing on software the land ERDAS cover IMAGE classification (Version map9.2). toThe calculate Iterative theSelf‐Organizing area of land Data cover Analysis types of each Technique Algorithm (ISODATA) [31,32], which is an unsupervised classification, was used to river basin.produce The resultsa yearly were land cover then classification divided by map the riverof Japan basin using area the to12 determinemonthly maximum the percentage NDVI of the area coveredimagery by datasets. each type Figure of the1 shows land an cover. example Table of the1 generated shows some land cover characteristics maps in 1996. of The the river land cover classes, suggestingbasin map was that superimposedCamellia japonicaon the landcommunity cover classification (LU4 map) was to calculate the largest the area land of land cover, cover occupying types of each river basin. The results were then divided by the river basin area to determine the about 29.2%, followed by beech type secondary vegetation (LU3, about 26.2%), beech type natural percentage of the area covered by each type of the land cover. Table 1 shows some characteristics of vegetation (LU2, about 20.4%), and plantation and field weed community (LU5, about 14.1%). It can the land cover classes, suggesting that Camellia japonica community (LU4) was the largest land cover, be foundoccupying from this about table 29.2%, that followed the maximum by beech type variation secondary in vegetation land cover (LU3, wasabout observed26.2%), beech for type beech type natural vegetationnatural vegetation (235.4%) (LU and2, about minimum 20.4%), and of plantation variation and was field observed weed community for Ccamellia (LU5, about japonica 14.1%).community (117.7%).It For can detecting be found from the this relationships table that the betweenmaximum pollutantvariation in loadland cover and was basin observed characteristics, for beech type the Pearson correlationnatural [33] wasvegetation conducted (235.4%) by usingand minimum MATLAB of forvariation the above-collected was observed for dataset. Ccamellia As forjaponica the pollutant community (117.7%). For detecting the relationships between pollutant load and basin characteristics, load per area,the Pearson we did correlation not find [33] awas significant conducted by correlation using MATLAB coefficient for the above between‐collected them dataset. and As the for five land cover classes.the pollutant Only forload beech per area, type we did natural not find vegetation a significant werecorrelation the coefficient correlation between coefficients them and of the TN, COD, and DO afive little land bit cover higher. classes. However, Only for beech significant type natural correlation vegetation were coefficients the correlation between coefficients pollutant of TN, loads and land coverCOD, were and found DO a little for bit land higher. cover However, of urban significant area, correlation beech type coefficients natural between vegetation, pollutant and loads beech type and land cover were found for land cover of urban area, beech type natural vegetation, and beech secondarytype vegetation. secondary vegetation. The urban The area urban is area highly is highly related related to to the the export export of of TN, TN, TP, TP, BOD, BOD, COD, COD, and and DO from riverDO basins. from river basins.

Figure 1. Generated land cover from National Oceanic and Atmospheric Administration (NOAA) Figure 1. AdvancedGenerated Very land High coverResolution from Radiometer National (AVHRR) Oceanic remote and sensing Atmospheric imagery (example Administration from 1996). (NOAA) Advanced Very High Resolution Radiometer (AVHRR) remote sensing imagery (example from 1996). Table 1. Statistical parameters of land cover (SD means standard deviation; CV means coefficient of variation).

Table 1. StatisticalVariables parameters Mean of land coverRange (SD meansSD standardCV% deviation;Mean Percent CV means coefﬁcient of variation). LU1 (km2) 222.3 4.0–1363.0 299.8 134.8 10.1% LU2 (km2) 450.2 <0.01–4721.0 1059.7 235.4 20.4% LU3 (km2) 579.7 <0.01–5365.0 1091.1 188.2 26.2% Variables Mean Range SD CV% Mean Percent LU4 (km2) 646.0 <0.01–2696.0 760.6 117.7 29.2% 22 LULU15 (km(km) ) 222.3311.0 4.0–1363.0<0.01–1653.0 464.9 299.8 149.5134.8 10.1%14.1% 2 Notes: AbbreviationsLU2 (km ) are 450.2as follows:<0.01–4721.0 LU1, urban area; LU1059.72, beech type235.4 natural vegetation;20.4% LU3, beech 2 type secondaryLU3 (km vegetation;) 579.7 LU4, Camellia<0.01–5365.0 japonica community;1091.1 LU5, plantation188.2 and field26.2% weed community. 2 LU4 (km ) 646.0 <0.01–2696.0 760.6 117.7 29.2% 2 LU5 (km ) 311.0 <0.01–1653.0 464.9 149.5 14.1%

Notes: Abbreviations are as follows: LU1, urban area; LU2, beech type natural vegetation; LU3, beech type secondary vegetation; LU4, Camellia japonica community; LU5, plantation and ﬁeld weed community. Water 2016, 8, 361 4 of 14

Water 2016, 8, 361 4 of 13 As for the hydrochemical datasets, we collected available long-term observation data of river discharge,As TN, for the TP, hydrochemical BOD, COD, datasets, and DO we concentration collected available for long 30 selected‐term observation river outlets data of (Figure river 2), whichdischarge, represented TN, TP, different BOD, COD, regions and DO of concentration Japan, for 11 for years 30 selected (1994–2004). river outlets Statistical (Figure 2), analysis which has beenrepresented conducted different for the collectedregions of monthly Japan, for hydrochemical 11 years (1994–2004). database Statistical from the analysis National has Land been with Waterconducted Information for the (http://www1.river.go.jp/) collected monthly hydrochemical monitoring database network from the [14 National]. Table2 Landshows with the Water statistical descriptionInformation of the (http://www1.river.go.jp/) collected water quality data.monitoring For all network 30 sites [14]. over Table Japan, 2 large shows variations the statistical of annual concentrationdescription of of TN, the TP, collected and BOD water were quality observed data. For between all 30 sites rivers, over with Japan, a smallest large variations coefficient of ofannual variation as 10.3%concentration for DO and of highestTN, TP, value and BOD as 162.6% were forobserved TP [14 between]. The large rivers, spatial with variation a smallest of coefficient TP concentration of variation as 10.3% for DO and highest value as 162.6% for TP [14]. The large spatial variation of TP are corresponding to the large differences in regional characteristics including population, topography, concentration are corresponding to the large differences in regional characteristics including geology,population, agricultural topography, activities, geology, and agricultural land use in activities, 30 river and basins. land use Furthermore, in 30 river basins. significant Furthermore, correlation coefficientssignificant from correlation the Pearson coefficients correlation from analysis the Pearson [32] forcorrelation detecting analysis the relationships [32] for detecting between the water pollutantrelationships loads and between land coverwater havepollutant been loads found and [ 14land]. cover have been found [14].

FigureFigure 2. 2.Location Location of of the the selectedselected 30 30 river river basins basins in inJapan. Japan.

Table 2. Statistical parameters of the studied water quality variables (SD means standard deviation; Table 2. Statistical parameters of the studied water quality variables (SD means standard deviation; CV means coefficient of variation). CV means coefﬁcient of variation). Variables Mean Range SD CV% VariablesTN (mg/L) Mean1.7 Range0.5–11.3 SD2.2 129.7 CV% TP (mg/L) 0.1 <0.01–0.8 0.2 162.6 TN (mg/L) 1.7 0.5–11.3 2.2 129.7 TPBOD (mg/L) (mg/L) 0.11.8 <0.01–0.80.2–14.2 2.5 0.2 137.9 162.6 BODCOD (mg/L) (mg/L) 1.82.9 0.2–14.20.8–9.5 1.9 2.5 63.5 137.9 CODDO (mg/L) (mg/L) 2.910.1 0.8–9.56.7–11.7 1.0 1.9 10.3 63.5 DO (mg/L) 10.1 6.7–11.7 1.0 10.3 For a clearer understanding of the spatial distribution of water pollution loads in Japan, the basin area and river mouth discharge for each catchment are listed in Table 3. The catchment areas which Forhave a an clearer area larger understanding than 2000 km of2 the are spatial highlighted distribution with shaded of water color. pollution We found loads that catchments in Japan, the2, 3, basin area and4, 5, 7, river 12, 15, mouth 16, 17, discharge 18, and 28 for are each relatively catchment large compared are listed to in others Table and3. Thealso have catchment large total areas river which havedischarge an area larger volumes, than both 2000 of km which2 are contribute highlighted to large with total shaded water color. pollution We foundloads. that catchments 2, 3, 4, 5, 7, 12, 15, 16, 17, 18, and 28 are relatively large compared to others and also have large total river discharge volumes, both of which contribute to large total water pollution loads. Water 2016, 8, 361 5 of 14

Table 3. Basin area and river mouth discharge for each catchment.

Catchment 1 2 3 4 5 6 7 8 9 10 Area (km2) 1078 10,617 2252 4198 4594 1189 6869 858 1539 1217 Q (m3/L) 3.1 163.2 56.6 186.8 235.0 60.6 367.7 86.3 9.3 11.6 Catchment 11 12 13 14 15 16 17 18 19 20 Area (km2) 237 3878 582 1277 4854 4762 3695 2069 831 361 Q (m3/L) 3.2 40.4 6.4 34.7 132.6 205.3 80.4 40.2 4.2 6.8 Catchment 21 22 23 24 25 26 27 28 29 30 Area (km2) 511 929 1202 557 650 1368 509 2164 510 922 Q (m3/L) 14.1 18.6 45.1 18.5 17.9 36.7 11.3 47.0 19.3 24.2

Among the collected hydrochemical datasets, the data from 1994 to 1999 (6 years) were used for model calibration. The remaining 5 years of data (2000–2004) were used for model validation. Annual water pollution load in rivers was estimated as the product of the annual average water quality concentration and the annual discharge at the outlet of each river. All the original water quality and river discharge data were measured by the National Land with Water Information (http://www1.river.go.jp/) monitoring network.

2.2. Estimating Water Poolution Loads from Different Land Cover Classifications by a Modified Empirical Model An empirical model based on the above data was used to simulate annual pollutant loads by linking load with different anthropogenic and natural variables [14,26]. To estimate annual pollutant loads from each land cover classification for each river basin, this model was modified in this study. The water pollution load exported from each river basin i can be presented as follows: L Q S i = k( i )n · ( i ) (2) Ai Ai Ai where Li (kg/year) is the total annual water pollution load exported from river basin i, which can 2 be calculated as the product of average water quality concentration and river discharge. Ai (km ) is 3 the area of river basin i, Qi (m /year) is the annual river discharge at the river outlet of river basin i, Si (kg/year) is the annual runoff load accumulated from different land covers in river basin i; k and n are empirical coefficients which are calibrated value by using optimization method or trial and error method. In this equation, the effect of basin area on water pollution load is removed with the division by basin area. The annual water pollution load accumulating in river basin i, i.e., Si, is expressed by:

5 ´ Si = ∑(LUj · Aij) (3) j where j (values 1–5) means each land use category of class 1 to class 5 shown in Figure1. 2 2 LUj (kg/km ·year) is the coefﬁcient named as runoff load factor of each category j, and A´ij (km ) is 2 the land cover area of each category j in river basin i. Here, LUj (kg/km ·year) represents the average value of the annual total water pollution loads accumulating in each category j in all of Japan. As shown in the above Equation (2), there are several parameters which should be estimated using the collected hydrochemical and land cover dataset. In this study, the simplex method (also known as the polytope method), which linearly adjusts model parameters to meet convergence criteria [34,35], was used to estimate k, n, and LUj (j = 1 to 5).

3. Results

3.1. Model Calibration (1994–1999) and Validation (2000–2004) We estimated annual discharged water pollution loads using the data from 1994 to 1999 for the 30 study basins. The observed and predicted water pollution loads were compared in the 30 river Water 2016, 8, 361 6 of 14

Water 2016, 8, 361 6 of 13 basins for the calibration period (1994–1999). The regression of predicted versus observed water pollutionpollution loads loads resulted resulted in in highhigh correlation coefficients: coefficients: R2 R=2 0.92,= 0.92, 0.72, 0.72, 0.90, 0.90, 0.98, 0.98,and 0.94 and for 0.94 TN, for TP, TN, TP,BOD, BOD, COD, COD, and and DO, DO, respectively, respectively, indicating indicating that that water water pollution pollution loads loads estimates estimates had had sufficient sufficient reproducibilityreproducibility for for the the calibration calibration periodperiod (Figure(Figure 33).). Furthermore, we we can can see see from from Table Table 44 that that the the calibratedcalibrated river river discharge discharge factor, factor, i.e., i.e., coefficient coefficient n in Equation (1), (1), has has the the smallest smallest value value for for DO DO and and largestlargest for for TN, TN, and and has has almost almost the the same same value value forfor BOD, COD, and and TP. TP.

FigureFigure 3. 3.Comparison Comparison of of observed observed and predicted predicted pollutant pollutant loads loads in in30 30river river basins basins in Japan in Japan for the for 6 the 6 yearyear calibrationcalibration period (1994–1999). (1994–1999). The The 1:1 1:1 line line is isplotted plotted for for comparison. comparison.

TableTable 4. 4.Estimated Estimated values of model parameters. parameters.

Parameter k n LU1 LU2 LU3 LU4 LU5 ParameterTN k0.09 n0.81 LU0.801 LU0.042 0.09LU 3 0.00LU 4 0.34 LU5 TNTP 0.090.17 0.810.44 0.800.36 0.83 0.04 0.91 0.09 0.52 0.00 0.76 0.34 TPBOD 0.170.14 0.440.73 0.360.68 0.29 0.83 0.32 0.91 0.15 0.52 0.28 0.76 BODCOD 0.140.19 0.730.70 0.681.50 0.97 0.29 0.76 0.32 0.55 0.15 0.50 0.28 COD 0.19 0.70 1.50 0.97 0.76 0.55 0.50 DO 1.62 0.34 0.00 115.73 33.40 47.92 71.08 DO 1.62 0.34 0.00 115.73 33.40 47.92 71.08

Water 2016, 8, 361 7 of 14 Water 2016, 8, 361 7 of 13

FiguresFigures3 and3 and4 show 4 show the the model model calibration calibration and and validation validation resultsresults withwith the comparison of of the the measuredmeasured and and validated validated annual annual water water pollution pollution loads.loads. The magnitude magnitude of of the the predicted predicted annual annual water water pollutionpollution loads loads closely closely followed followed the the measured measured datadata in most of of the the studied studied rivers. rivers. We We found found good good simulationsimulation results results for for TN, TN, BOD, BOD, COD, COD, and and DO DO when when comparing comparing the ﬁguresthe figures with with the calibration the calibration results. Theresults. simulation The simulation of TP showed of TP the showed smallest the correlationsmallest correlation coefﬁcient coefficient between between the observed the observed and modeled and TPmodeled load. In termsTP load. of In the terms description of the description itself, it appears itself, it that appears the model that the overestimates model overestimates TP loads TP from loads some basinsfrom in some which basins the actualin which TP the is veryactual low. TP is very low.

FigureFigure 4. 4.Comparison Comparison of of observed observed and and predicted predicted water pollution loads loads in in 30 30 river river basins basins in in Japan Japan for for thethe 5 year 5 year validation validation period period (2000–2004). (2000–2004). The The 1:11:1 lineline is plotted for for comparison. comparison.

The above calibration and validation were conducted at the whole‐of‐catchment scale, which The above calibration and validation were conducted at the whole-of-catchment scale, which includes all the land cover types. To evaluate the accuracy of the predictions at the scale of individual includes all the land cover types. To evaluate the accuracy of the predictions at the scale of individual land cover types, the water pollution loads were predicted from different land cover type. Table 5 land cover types, the water pollution loads were predicted from different land cover type. Table5 shows the R2 coefficients for predicting TN, TP, BOD, COD, and DO by using each land cover type. shows the R2 coefﬁcients for predicting TN, TP, BOD, COD, and DO by using each land cover type. From this table, it can be found that the most accurate prediction of water pollution loads is that Frompredicted this table, by using it can land be cover found of thaturban the area. most The accurate second accurate prediction prediction of water of water pollution pollution loads loads is that predictedis that predicted by using by land using cover land of urbancover of area. beech The type second natural accurate vegetation. prediction The least of water accurate pollution predictions loads is thatof predicted water pollution by using loads land are cover those of beechpredicted type by natural land cover vegetation. of Camellia The least japonica accurate community, predictions and of waterPlantation pollution (Cryptomeria loads are thosejaponica predicted and Cryptomeria by land) cover and field of Camellia weed community. japonica community, and Plantation (Cryptomeria japonica and Cryptomeria) and ﬁeld weed community.

Water 2016, 8, 361 8 of 14

Table 5. Accuracy of the predictions at the scale of individual land cover types. Water 2016, 8, 361 8 of 13

Table 5. Accuracy of theLU predictions1 LU2 at theLU scale3 of individualLU4 LU land5 cover types. TN 0.71 0.35 0.33 0.00 0.00 LU1 LU2 LU3 LU4 LU5 TP 0.43 0.34 0.44 0.00 0.00 BODTN 0.740.71 0.660.35 0.600.33 0.000.00 0.000.00 CODTP 0.830.43 0.780.34 0.660.44 0.000.00 0.000.00 BODDO 0.790.74 0.720.66 0.680.60 0.000.00 0.000.00 COD 0.83 0.78 0.66 0.00 0.00 DO 0.79 0.72 0.68 0.00 0.00 3.2. Spatial Comparison of Water Pollution Loads in the 30 Rivers Considering3.2. Spatial Comparison the water of Water pollution Pollution loads Loads of 1996 in the from 30 Rivers our previous study [24], Figure5 shows the water pollutionConsidering loads the associated water pollution with loads different of 1996 land from cover our previous classifications study [24], in the Figure 30 catchments5 shows the for 1996,water as an pollution example. loads There associated was good with agreement different between land cover the classifications observed and in modeled the 30 catchments water pollution for loads1996, for TN as an and example. COD. ForThere most was of good the agreement catchments, between the results the observed are good and at modeled simulating water water pollution pollution loadsloads for BOD for TN and and DO. COD. As For for most TP, of there the catchments, are differences the results between are good the at observed simulating and water modeled pollution load. Moreover,loads itfor can BOD be and found DO. from As for the TP, figure there that are urban differences areas mainlybetween impacted the observed TN, and BOD, modeled and COD load. loads. Moreover, it can be found from the figure that urban areas mainly impacted TN, BOD, and COD The effect of urban areas on TP and DO was very small. Additionally, it is interesting to find that loads. The effect of urban areas on TP and DO was very small. Additionally, it is interesting to find the largethat the water large pollution water pollution loads ofloads TN, of TP, TN, BOD, TP, BOD, COD, COD, and and DO DO mostly mostly belong belong to to thethe river basin basin in the northernin the northern part of part Japan, of Japan, namely namely Kanto Kanto Region Region (from (from catchment catchment 1 to to 9). 9). The The second second highest highest water water pollutionpollution loads loads are are found found in centralin central part part of of Japan, Japan, namelynamely Kansai Kansai Region Region (from (from catchment catchment 10 to 10 17). to 17). The waterThe water pollution pollution loads loads are mostlyare mostly very very small small in in the the western western part part of of Japan Japan (from (from catchment catchment 18 to 30). It can30). be It also can foundbe also found that the that impact the impact of urban of urban area area on on water water pollution pollution loads loads is is large large in inboth both Kanto Kanto RegionRegion and and Kansai Kansai Region, Region, in whichin which the the economic economic is is wellwell developed and and population population density density is much is much higherhigher than than other other parts parts of Japan.of Japan.

FigureFigure 5. Water 5. Water pollution pollution loads loads from from different different land land covers covers in 30 30 Japanese Japanese catchments catchments (1996). (1996). The The lines lines represent the observed water pollution loads; the colored bars represent the water pollution represent the observed water pollution loads; the colored bars represent the water pollution loads loads associated with the different land covers. associated with the different land covers.

Water 2016, 8, 361 9 of 14 Water 2016, 8, 361 9 of 13

3.3. Long Long-Term‐Term VariationVariation inin WaterWater PollutionPollution LoadsLoads EstimatedEstimated fromfrom DifferentDifferent Sources Sources As evident from the calibration and validation results, the model exhibited a good associationassociation between measured measured and and predicted predicted water water pollution pollution load load values. values. Based Based on the on above the above simulation, simulation, seven sevenriver catchments river catchments including including Koyoshi, Koyoshi, Sendai, Sendai, Oze, Oze, Oita, Oita, Kitakami, Kitakami, Kuji, Kuji, and and Abe, Abe, which which are representative ofof the the 3 main3 main regions regions (north, (north, central, central, and western), and western), were selected were selected to take asto the take examples as the toexamples elaborate to long-termelaborate long variation‐term invariation water pollution in water loadspollution estimated loads fromestimated different from sources different from sources 1994 tofrom 2004. 1994 Contributions to 2004. Contributions of different of sources different to sources major water to major quality water variables quality canvariables be then can identiﬁed be then foridentified the investigated for the investigated river basins. river Figures basins.6 Figuresand7 show 6 and examples 7 show examples of TN and of COD TN and for COD seven for typical seven riverstypical for rivers which for long-term which long observation‐term observation data were data available. were available. The yellow The line yellow represents line represents the observed the waterobserved pollution water loads.pollution The loads. vertical The bars vertical indicate bars the indicate water pollution the water loads pollution from differentloads from land different covers. Fromland covers. Figure 6From, it can Figure be found 6, it can that be urban found areas that hadurban a large areas impact had a large on TN impact export on into TN rivers. export Itinto is probablyrivers. It is because probably there because are many there nitrogen are many sources nitrogen such sources as residential such as fertilizers, residential septic fertilizers, systems, septic and domesticsystems, and animal domestic waste animal in urban waste areas. in Even urban though areas. theEven urban though area the ratios urban were area low ratios for thewere Kitakami, low for Kuji,the Kitakami, Oita, Sendai, Kuji, Oita, and KoyoshiSendai, and Rivers, Koyoshi the waterRivers, pollution the water load pollution from urbanload from areas urban was areas still very was 3 high.still very Therefore, high. beechTherefore, type secondarybeech type vegetation secondary (LU vegetation3) and plantation (LU ) ﬁeldand weedplantation community field (LUweed5) 5 contributedcommunity strongly(LU ) contributed to the TN loadsstrongly in the to rivers,the TN whereas loads in beech the rivers, type natural whereas vegetation beech type (LU 2natural) made avegetation small contribution. (LU2) made Discharge a small contribution. from the Abe Discharge River increased from the dramatically Abe River increased in 2003, from dramatically 10 m3/s toin 1462003, m from3/s, and 10 m TN3/s exportto 146 m increased3/s, and TN accordingly. export increased It also canaccordingly. be found It that also a smallcan be contribution found that a ofsmall TN mightcontribution make aof large TN might difference make to a riverine large difference TN in small to riverine catchments. TN in small catchments.

Figure 6. Total nitrogennitrogen (TN)(TN) loadsloads associatedassociated with different land covers in 3030 JapaneseJapanese catchmentscatchments from 1994 to 2004.

From Figure 7, it can be found that urban areas also had a high impact on COD export into rivers. From Figure7, it can be found that urban areas also had a high impact on COD export into rivers. As with the COD results, the pollutant load from urban areas (LU1) was high for the Kitakami, Kuji, As with the COD results, the pollutant load from urban areas (LU1) was high for the Kitakami, Kuji, Oita, and Sendai Rivers. Land cover of beech type secondary vegetation (LU3) and Camellia japonica Oita, and Sendai Rivers. Land cover of beech type secondary vegetation (LU3) and Camellia japonica community (LU4) had a large effect on COD export into rivers. The impact from land cover beech community (LU4) had a large effect on COD export into rivers. The impact from land cover beech type type natural vegetation (LU2) was small, except for its effect on the Koyoshi River. As mentioned naturalabove, discharge vegetation from (LU2 )the was Abe small, River except increased for its effect suddenly on the in Koyoshi 2003, and River. COD As mentionedexport increased above, dischargeaccordingly. from the Abe River increased suddenly in 2003, and COD export increased accordingly.

Water 2016, 8, 361 10 of 14 Water 2016, 8, 361 10 of 13

TheThe same same analysis analysis was was conducted conducted for for the the other other water water pollutant pollutant indicators. indicators. We We found found that beech typetype natural natural vegetation vegetation (LU (LU2)2 )and and beech beech type type secondary secondary vegetation vegetation (LU (LU3) had3) had a larger a larger impact impact on TP on exportTP export into intoall rivers. all rivers. Urban Urban areas areas (LU1) (LU had1) a had very a small very smallimpact impact on DO on export DO export into rivers, into rivers,but a relativelybut a relatively large impact large impact on BOD on export. BOD export.

FigureFigure 7. 7. ChemicalChemical oxygen oxygen demand demand (COD) (COD) loads loads associated associated with with different different land land covers covers in in 30 30 Japanese Japanese catchmentscatchments from from 1994 to 2004.

4.4. Discussion Discussion

4.1.4.1. Model Model Advantage Advantage Compared Compared with Conventional Method AsAs stated stated in in the the above, above, Figures Figures 3 and4 4 showed showed good good agreement agreement between between modeled modeled and and measured measured water pollutionpollution loadsloads for for the the calibration calibration and and validation validation periods. periods. In the In calibrationthe calibration phase, phase, attempts attempts were weremade made to minimize to minimize the rootthe root mean mean square square error error and and obtain obtain R 2R2values values closest closest toto a value of of unity. unity. However,However, except for for thethe BOD BOD validation validation period, period, the the model model underestimated underestimated pollutant pollutant fluxes. fluxes. This This may may bebe partly partly due due to to the the model’s model’s inability inability to to capture capture the the effects effects of of consecutive consecutive heavy heavy rainfalls rainfalls that that caused caused floodsfloods throughout throughout Japan. Japan. A Astatistical statistical evaluation evaluation of the of measured the measured and observed and observed water water pollution pollution loads yieldedloads yielded high R high2 values R2 valuesclose to close 1.0, with to 1.0, a withhigh slope a high close slope to close 1.0 for to all 1.0 pollution for all pollution indicators, indicators, except TP.except Therefore, TP. Therefore, using this using method, this method, it was itpossible was possible to estimate to estimate the annual the annual total water total water pollution pollution load dischargedload discharged from fromthe river the riverbasins basins in Japan in Japan and identify and identify water water pollution pollution load changes load changes compared compared with conventionalwith conventional methods methods that estimate that estimate annual annual total water total pollution water pollution loads based loads on based the sum on the of suma basic of unit,a basic which unit, whichis the isdischarge the discharge rate of rate total of totalpollutant pollutant load load to each to each land land cover cover category category in in a a basin. basin. DeterminingDetermining a a basic basic unit unit for for each land cover category is is difficult, difficult, because there are many databases supplyingsupplying basic basic units units of of total pollutant loads loads evaluated in in various river river basins, gained by different methods,methods, and covering covering different sites and seasons. Compared Compared with with the the conventional conventional method method (such (such as as water quality basic unit method), it was easier to assess annual pollutant loads in river basins using thethe model proposed in this study.

Water 2016, 8, 361 11 of 14

4.2. Problems of Water Pollution Loads Since Figures3 and4 are the results of multiple years of different basins, land cover in those basins for different years is changing. The precipitation and TP leaching may rates may have differed from the rates at which the models were calibrated. Another reason may be that diffuse phosphorus source pollution is associated with agriculture because of the potential movement of fertilizers from the land surface directly into rivers and streams or through permeable soil to groundwater. For example, Camellia japonica community should be fertilized in late winter or very early in the spring when new growth begins, using a slow-release. The main mechanism of phosphorus loss from agricultural soil is through erosion, while leaching to groundwater and phosphorus loss due to landslide and debris flow cannot be considered in this method. Therefore, the effects of landscape management and of specific pollution practices on TP were difficult to define or quantify in the current study. Figures6 and7 give us a brief image of the trend of water pollution loads and indicate that in most of the rivers in Japan, water quality management has proven successful in improvement of BOD and COD through stringent discharge regulation, development of sewage systems, etc. However, the risk of water pollution still exists in some rivers. For example, in the Oita River the water pollution loads of both TN and COD are not decreasing, as shown in Figure6. In some water bodies, the environment status of DO has been worsening rather than improving [36].

4.3. Model Limitation On the other hand, the approach used in this study has some limitations. The most important limitation is the uncertainty in parameter calibration. The polytope method linearly adjusted model parameters to meet convergence criteria for the estimation of k, n, and LUj. The parameter values shown in Table4 may change, however, as characteristics of the river basins change. Additionally, we found that the TN, BOD, and COD runoff load factors from urban areas were higher than those of other land cover types (beech type natural vegetation, beech type secondary vegetation, Camellia japonica community, and plantation field weed community). In another word, these results indicate higher levels of TN, BOD, and COD from urban areas. In urban areas, the nitrogen concentration of rain is higher than for other natural land covers due to various factors such as atmospheric pollution [26]. Moreover, total water pollution loads are discharged from urban areas, unlike in forested areas, where the total water pollution loads cannot be discharged easily due to forest uptake and soil absorption. In addition, the BOD and COD concentrations of urban area river waters were higher than those of other natural land cover areas due to human activities. Furthermore, the results of river discharge factor, i.e., coefficient n in Equation (1), differed slightly from those of He et al. [14], who used the hydrochemical dataset from 1996. In the present study, we used the average of values from 1994 to 2004 for all hydrochemical datasets (river discharge, TN, TP, BOD, COD, and DO). This means the size of the calibration dataset also has impact on the parameter calibration. In addition, n was small, which does not mean that the correlation between DO load and river discharge was weak; rather, the value of the coefficient n was attributable to the magnitude of the data, and BOD load ranged from several hundreds to tens of thousands, and TP load ranged from zero to several hundred. Furthermore, the parameterized water quality model in this study simplified the complex surface water hydraulics, chemical process, and transport processes. It works on yearly time scale simulation without providing the seasonal change of water quality in rivers and the impact of river discharge on water quality concentration. Therefore, the process-based water quality model can be more useful in case the detailed hydrochemical process needs to be accounted for. In addition, the NOAA–NDVI generated land cover classification map decides the reliability of the water quality model. However, the relationships between NDVI and water quality are likely to change with atmospheric CO2 concentrations, temperature, precipitation, and so on in the future. This can be more clearly demonstrated in process-based water quality models. Even though there are various limitations with this approach, this study gives a simple and efficient way to source apportionment of annual water pollution loads in river basins by remote-sensed land cover classification. Water 2016, 8, 361 12 of 14

5. Conclusions The major objective of this study has been to apply a simple empirical water quality model for examining the water pollution loads from land cover classifications. The authors analyzed the relationship between long-term observation hydrochemical data and land cover classifications estimated from NOAA AVHRR imagery. Statistical analysis revealed large spatial variations in pollutant concentrations of TN, TP, BOD, river discharge, and land use. Urban areas were highly associated with the export of TN, BOD, and COD from river basins. This study indicated that urban and forest land may be the dominant sources of water pollution loads. The simple empirical water quality model was modified and applied successfully in 30 rivers in Japan. Calibration and validation results also indicated that the proposed simulation technique was useful for predicting water pollution loads in the river basins. As a conclusion, this study illustrated the usefulness of using land cover classification and an efficient empirical model for identification of pollution sources and understanding variations in water quality for effective water quality management.

Acknowledgments: This study was sponsored by the National Natural Science Foundation of China (No. 41471460, 41501552), the “One Hundred Talents Program” of the Chinese Academy of Sciences, and the Asia-Pacific Network for Global Change Research (APN) “CAF2014-RR06(ARCP)-NMY-Wang: Integrated Flood Modelling and Pre-disaster Loss Estimation in Asian Countries”. The authors also thank the Sawada and Takeuchi Laboratory, Institute of Industrial Science, University of Tokyo for providing the NOAA AVHRR dataset. Author Contributions: Yi Wang and Bin He had the original idea for the study and, with all co-authors carried out the design. Yi Wang drafted the manuscript, which was submitted by Weili Duan. All authors read and approved the final manuscript. Conflicts of Interest: The authors declare no conflict of interest.

References

1. Changnon, S.A.; Demissie, M. Detection of changes in streamflow and floods resulting from climate fluctuations and land use-drainage changes. Clim. Chang. 1996, 32, 411–421. [CrossRef] 2. He, B.; Wang, Y.; Takase, K.; Mouri, G.; Razafindrabe, B.H. Estimating land use impacts on regional scale urban water balance and groundwater recharge. Water Resour. Manag. 2009, 23, 1863–1873. [CrossRef] 3. Duan, W.; He, B.; Nover, D.; Yang, G.; Chen, W.; Meng, H.; Zou, S.; Liu, C. Water Quality Assessment and Pollution Source Identification of the Eastern Poyang Lake Basin Using Multivariate Statistical Methods. Sustainability 2016, 8, 133. [CrossRef] 4. Pearce, N.J.; Yates, A.G. Agricultural Best Management Practice Abundance and Location does not Influence Stream Ecosystem Function or Water Quality in the Summer Season. Water 2015, 7, 6861–6876. [CrossRef] 5. Ngoye, E.; Machiwa, J.F. The influence of land-use patterns in the Ruvu river watershed on water quality in the river system. Phys. Chem. Earth A/B/C 2004, 29, 1161–1166. [CrossRef] 6. Sliva, L.; Williams, D.D. Buffer zone versus whole catchment approaches to studying land use impact on river water quality. Water Res. 2001, 35, 3462–3472. [CrossRef] 7. Duan, W.; Takara, K.; He, B.; Luo, P.; Nover, D.; Yamashiki, Y. Spatial and temporal trends in estimates of nutrient and suspended sediment loads in the Ishikari River, Japan, 1985 to 2010. Sci. Total Environ. 2013, 461, 499–508. [CrossRef][PubMed] 8. Duan, W.; He, B.; Takara, K.; Luo, P.; Nover, D.; Sahu, N.; Yamashiki, Y. Spatiotemporal evaluation of water quality incidents in Japan between 1996 and 2007. Chemosphere 2013, 93, 946–953. [CrossRef][PubMed] 9. Seitzinger, S.P.; Kroeze, C. Global distribution of nitrous oxide production and N inputs in freshwater and coastal marine ecosystems. Glob. Biogeochem. Cycles 1998, 12, 93–113. [CrossRef] 10. Strayer, D.L.; Beighley, R.E.; Thompson, L.C.; Brooks, S.; Nilsson, C.; Pinay, G.; Naiman, R.J. Effects of land cover on stream ecosystems: Roles of empirical models and scaling issues. Ecosystems 2003, 6, 407–423. [CrossRef] 11. Turner, R.E.; Rabalais, N.N. Linking landscape and water quality in the Mississippi River basin for 200 years. Bioscience 2003, 53, 563–572. [CrossRef] 12. Smith, S.V.; Swaney, D.P.; Buddemeier, R.W.; Scarsbrook, M.R.; Weatherhead, M.A.; Humborg, C.; Eriksson, H.; Hannerz, F. River nutrient loads and catchment size. Biogeochemistry 2005, 75, 83–107. [CrossRef] Water 2016, 8, 361 13 of 14

13. Paul, M.J.; Meyer, J.L. Streams in the urban landscape. In Urban Ecology; Springer: Boston, MA, USA, 2008; pp. 207–231. 14. He, B.; Oki, T.; Kanae, S.; Mouri, G.; Kodama, K.; Komori, D.; Seto, S. Integrated biogeochemical modelling of nitrogen load from anthropogenic and natural sources in Japan. Ecol. Model. 2009, 220, 2325–2334. [CrossRef] 15. He, B.; Kanae, S.; Oki, T.; Hirabayashi, Y.; Yamashiki, Y.; Takara, K. Assessment of global nitrogen pollution in rivers using an integrated biogeochemical modeling framework. Water Res. 2011, 45, 2573–2586. [CrossRef] [PubMed] 16. Carpenter, S.R.; Caraco, N.F.; Correll, D.L.; Howarth, R.W.; Sharpley, A.N.; Smith, V.H. Nonpoint pollution of surface waters with phosphorus and nitrogen. Ecol. Appl. 1998, 8, 559–568. [CrossRef] 17. Boyer, E.W.; Goodale, C.L.; Jaworski, N.A.; Howarth, R.W. Anthropogenic nitrogen sources and relationships to riverine nitrogen export in the northeastern USA. In The Nitrogen Cycle at Regional to Global Scales; Springer: Boston, MA, USA, 2002; pp. 137–169. 18. Sivertun, Å.; Prange, L. Non-point source critical area analysis in the Gisselö watershed using GIS. Environ. Model. Softw. 2003, 18, 887–898. [CrossRef] 19. He, B.; Oki, T.; Sun, F.; Komori, D.; Kanae, S.; Wang, Y.; Kim, H.; Yamazaki, D. Estimating monthly total nitrogen concentration in streams by using artiﬁcial neural network. J. Environ. Manag. 2011, 92, 172–177. [CrossRef][PubMed] 20. Ding, J.; Jiang, Y.; Fu, L.; Liu, Q.; Peng, Q.; Kang, M. Impacts of Land Use on Surface Water Quality in a Subtropical River Basin: A Case Study of the Dongjiang River Basin, Southeastern China. Water 2015, 7, 4427–4445. [CrossRef] 21. Tachibana, H.; Yamamoto, K.; Yoshizawa, K.; Magara, Y. Non-point pollution of Ishikari river, Hokkaido, Japan. Water Sci. Technol. 2001, 44, 1–8. [PubMed] 22. Kimura, S.D. Creation of an Eco-Balance Model to Assess Environmental Risks Caused by Nitrogen Load in a Basin-Agroecosystem. Ph.D. Thesis, Hokkaido University, Hokkaido, Japan, 25 March 2005. 23. Kimura, S.D.; Hatano, R. An eco-balance approach to the evaluation of historical changes in nitrogen loads at a regional scale. Agric. Syst. 2007, 94, 165–176. [CrossRef] 24. Shrestha, S.; Kazama, F. Assessment of surface water quality using multivariate statistical techniques: A case study of the Fuji river basin, Japan. Environ. Modell. Softw. 2007, 22, 464–475. [CrossRef] 25. Duan, W.L.; He, B.; Takara, K.; Luo, P.P.; Nover, D.; Hu, M.C. Modeling suspended sediment sources and transport in the Ishikari River basin, Japan, using SPARROW. Hydrol. Earth Syst. Sci. 2015, 19, 1293–1306. [CrossRef] 26. Oki, K.; Yasuoka, Y. Mapping the potential annual total nitrogen load in the river basins of Japan with remotely sensed imagery. Remote Sens. Environ. 2008, 112, 3091–3098. [CrossRef] 27. He, B.; Oki, K.; Wang, Y.; Oki, T. Using remotely sensed imagery to estimate potential annual pollutant loads in river basins. Water Sci. Technol. 2009, 60, 2009–2015. [CrossRef][PubMed] 28. Sawada, H.; Takeuchi, W. Sawada and Takeuchi Laboratory. Ph.D. Thesis, Institute of Industrial Science, The University of Tokyo, Tokyo, Japan, 2009. 29. Shimoda, H.; Fukue, K.; Cho, K.; Matsuoka, R.; Hashimoto, T.; Nemoto, T.; Tokunaga, M.; Tanba, S.; Takagi, M. Development of a software package for ADEOS and NOAA data analysis. In Proceedings of the IEEE Geoscience and Remote Sensing Symposium, Seattle, WA, USA, 6–10 July 1998; Volume 2, pp. 674–676. 30. Rouse, J.W.; Haas, R.H.; Schell, J.A.; Deering, D.W. Monitoring vegetation systems in the great plains with ERTS. In Proceedings of the Third Earth Resources Technology Satellite-1 Symposium, NASA SP-351, Washington, DC, USA, 14 December 1974; pp. 309–317. 31. Dubes, R.C.; Jain, A.K. Algorithms for Clustering Data Prentice-Hall; Prentice-Hall, Inc.: Upper Saddle River, NJ, USA, 1988. 32. Jensen, J.R. Introductory Digital Image Processing: A Remote Sensing Perspective; Prentice Hall: Old Tappan, NJ, USA, 1986. 33. Bolboaca, S.; Jäntschi, L. Pearson versus Spearman, Kendall’s tau correlation analysis on structure-activity relationships of biologic active compounds. Leonardo J. Sci. 2006, 5, 179–200. Water 2016, 8, 361 14 of 14

34. Nelder, J.A.; Mead, R. A simplex method for function minimization. Comput. J. 1965, 7, 308–313. [CrossRef] 35. Press, W.H. Numerical Recipes 3rd Edition: The Art of Scientiﬁc Computing; Cambridge University Press: Cambridge, UK, 2007. 36. Ando, H.; Kashiwagi, N.; Ninomiya, K.; Ogura, H.; Kawai, T. Changes in the State of Water Pollution in Tokyo Bay since 1980-Trend analysis of water quality using monitoring data obtained by Local Governments. Annu. Rep. Tokyo Metrop. Res. Inst. Environ. Prot. 2005, 2005, 141–150.

© 2016 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC-BY) license (http://creativecommons.org/licenses/by/4.0/).